Amiodarone and the risk of cancer: A nationwide population‐based study

I read with interest the population-based cohort secondary analysis of the Taiwan National Health Insurance Research database.1 This is a great first step toward hypothesis generation and I commend the authors on this work. The statistically significant increase in cancer incidence in the subgroup of male patients and those receiving higher cumulative defined daily doses of amiodarone who had been receiving long-term treatment with amiodarone is worth further comment. The investigators subdivide the study cohort into many other subgroups, including those based on duration of treatment, cancer site, and age group (as shown in Tables 2 and 3).1 Finding statistical significance in one subgroup among many needs further analysis.2, 3 If this subgroup analysis was decided a priori, I would suggest that the investigators comment on the value of correcting for multiple testing.3 This is relevant because there is a 40% chance of finding at least 1 statistically significant false-positive result when performing 10 comparisons of the primary endpoint.4 To further elaborate on this point, any cohort could be “sliced” in so many halves based on sex, race, ethnicity, religion, nationality, study center, age group, or any other distinction. The odds are that a subgroup will be above that neutrality line by chance alone, and the correspondent other category will be on the other side of the neutrality line, so that their effects cancel each other. In addition, if the overall cohort study results are neutral (ie, the null hypothesis is true), but a subgroup of patients is found to have an increased risk, the correlate of this is that the opposite correspondent subgroup is “protected” by that study intervention (in this study example, amiodarone protects women against the risk of cancer, which is not necessarily true). Addressing each of these questions would minimize the inherent potential of bias when performing subgroup analyses. I again congratulate the investigators on a well-done study. The author made no disclosures. Wassim H. Fares, MD, MSc Pulmonary Vascular Disease ProgramPulmonary, Critical Care and Sleep Medicine SectionDepartment of Internal MedicineYale School of MedicineNew Haven, Connecticut